Removal of small trihalomethane precursors from aqueous solution by nanofiltration

(1)

Removal of small trihalomethane precursors from

aqueous solution by nanofiltration

Yi-Li Lin

a

, Pen-Chi Chiang

a,∗,1

, E.-E. Chang

b,2

a_{Graduate Institute of Environmental Engineering, National Taiwan University, 71 Chou-Shan Road, Taipei, Taiwan} b_{Department of Biochemistry, Taipei Medical University, 250 Wu-Hsing Street, Taipei 110, Taiwan}

Received 16 June 2006; received in revised form 20 November 2006; accepted 21 November 2006 Available online 30 November 2006

Abstract

The removal of small trihalomethane precursors (THMPs) from aqueous solution by two commercial nanofiltration membranes (NF70 and NF270) was investigated. Resorcinol, phloroglucinol, and 3-hydroxybenzoic acid were selected as model compounds of small THMPs, while tannic acid was chosen as a medium molecular disinfection by-product (DBP) precursor for comparison. The performance of nanofiltration membranes were evaluated by introducing polyethylene glycol (PEG) solutions and uncharged saccharides to estimate molecular weight cut-off (MWCO) and membrane pore radii, respectively. The streaming potential was measured to estimate the membrane surface charge at different pH values, which reveals that the NF270 membrane is more pH-sensitive than the NF70 membrane. The rejections of the above selected THMPs were assessed under various pH values, and the removal efficiencies of THMPs for both membranes at high pH values are reasonably well. Charge exclusion is the prevailing mechanism for the selected small model compounds retended by the negatively charged nanofiltration membranes, while size exclusion and adsorption are controlled mechanisms but not sufficient for the rejection of unionized small organic molecules. In general, the NF270 membrane exhibits the superior permeation rate value, which takes an advantage over the NF70 membrane from the aspect of energy conservation.

Keywords: Nanofiltration; Membrane; Trihalomethane precursor; Disinfection by-product; Drinking water production

1. Introduction

Trihalomethanes (THMs) are the major disinfection by-products (DBPs) but not naturally present in ground or surface waters. Certain organic compounds such as humic acid and ful-vic acid called THM precursors (THMPs) are able to form THMs when reacting with chlorine in the disinfection process. Previ-ous studies have demonstrated that THMs cause liver, kidney or central nervous system problems[1–2]. Therefore, the max-imum contaminant levels (MCL) of THMs were promulgated by many nations to protect public health[3–7]. US EPA is even adopting more stringent MCLs of THMs from 80 to 40␮g/L in the stage 2 disinfectants/DBP rule[8].

∗_{Corresponding author. Tel.: +886 2 2362 2510; fax: +886 2 2366 1642.}

E-mail addresses:f88541104@ntu.edu.tw(Y.-L. Lin),

pcchiang@ntu.edu.tw(P.-C. Chiang),eechang@tmu.edu.tw(E.-E. Chang). 1 _{Tel.: +886 2 2362 2510; fax: +886 2 2366 1642.}

2 _{Tel.: +886 2 2736 9236; fax: +886 2 2736 9236.}

As chlorine is a widely used disinfectant, the reduction of THMs formation potential (THMFP) is very important. Although the THMs can be removed by introducing the advanced treatment process, it is not feasible from the eco-nomic point of view. The usage of substituting disinfectant such as chloramine has raised attention of the formation of N-nitrosodimethylamine (NDMA) by-product, which is an emerging water contaminant and is a probable human carcino-gen classified by the US EPA[9]. On the other hand, removing THMPs before chlorination is a more effective and economic THMs control technology than the previous methods.

It is well known that the traditional coagulation process in a water treatment plant is not effective in removing small THMPs, while the membrane process has been proven to be effective[10–14]. Nanofiltration (NF), with molecular-weight-cutoff (MWCO) between the range of ultrafiltration (UF) and reverse osmosis (RO), was found to be more cost effective than RO due to lower operating pressure and has higher rejection efficiencies for small organics and multivalent ions than UF. Therefore, in various applications, RO was replaced with NF

(2)

for a higher membrane flux and significantly better permeate water quality. Many researchers have shown that NF membrane technology provides very high removal of natural organic matter (NOM) in drinking water treatment[10–14]. Owen et al.[11]

proposed that NF with a relatively lower MWCO of 400–800 Da is effective in controlling the formations of DBPs. Retentions of various micropollutants by NF membranes were also reported by several researchers[12–14].

NF now is competitive with other NOM removal technolo-gies such as conventional clarification and GAC adsorption because of the following advantages: simplicity of operation, development of higher flux membrane with low fouling poten-tials, and lower membrane costs[15]. Some researchers have indicated that membrane costs are comparable to or lower than conventional treatment for small systems of <20,000 m3/day, or 5 mg/day (million gallons per day)[15–16]. However, there are still a lot of unknowns in the removal mechanisms by NF mem-branes because the rejections of solutes by NF membrane are rather complicated. Membrane properties such as hydrophobic-ity, membrane charge, membrane pore size, potential for fouling, resistance to temperature, retention properties and permeability dramatically affect the filtration process. Solute properties such as dipole moment and hydrophibicity also affect the separation efficiency by adsorbing or interacting with membrane surfaces

[17–18].

Two main solute retention mechanisms such as size (sieving or steric interaction) and charge (Donnan) effects are gener-ally discussed in the area of membrane science and technology. Size exclusion is the main separation mechanism for uncharged solutes based on the relationship between the membrane pore size and permeating molecule size, while charge exclusion is caused by electrostatic interactions between the charged per-meating species and the charged membrane surface. However, the contributions of these two basic mechanisms in the NF pro-cess cannot be assessed simply and accurately because of the uncertainty of involving parameters. The NF separation per-formance is usually evaluated on a case by case basis, and many researchers describe the solute transport mechanisms with numerous kinds of models, such as the Spiegler–Kedem model

[19], Teorell–Mayer–Sievers and Steric Hindrance Pore (SHP) model[20], and the hindrance model[21].

Most organic matters that are responsible for major DBP precursors in Taiwan source water are small to medium com-pounds, with a molecular weight near or less than 1 kDa[22–24], which was consistent with the findings suggested by Amy et al. [25] that the majority of THMPs are present in apparent molecular weight (AMW) fractions less than l kDa. Evidence shows that resorcinol structures in aquatic humic materials are the major THMPs in colored waters [26], and the amount of chloroform that occurred by substituted phenols is affected by the position of chloror or hydroxyl atoms in the phenolic com-pound[1]. Several studies also suggest that aliphatic carboxylic acids, hydroxybenzoic acids, phenols and pyrrole derivatives are reactive substrates of organic precursors for THM formation

[27–28]. Resorcinol is known to affect the hormone production of the human thyroid. The hazardous score of resorcinol is 23.9 in accordance to the Indiana Relative Chemical Hazard Score

(IRCHS) database[29], which is considered highly hazardous. Since the very limited study on small aromatic DBP precur-sors with different functional groups and their treatabilities by nanofiltration, the three small organic compounds (resorcinol, phloroglucinol, and 3-hydrobenzoic acid) with different func-tional groups and high THMFP were selected to investigate their respective performance and removal mechanisms treated by the NF membranes.

The objectives of this research were to assess the character-istics of two commercial nanofiltration membranes (NF70 and NF270) and then to determine the efficiencies of these two mem-branes for removing small THMPs. Tannic acid was chosen as a major medium molecular DBP precursor for comparison. The performances of these two NF membranes were examined with uncharged solutes at the beginning of the experiments. The mem-brane surface charge was estimated by measuring the streaming potential. The effects of solute types on filtration and the reten-tion mechanisms at various levels of pH were also investigated.

2. Materials and methods

2.1. Membranes and model compounds

In this study, two types of commercial thin-film polyamide NF membranes, NF70 and NF270, produced by Dow-FilmTec were examined. Table 1 summarizes the characteristics of NF70 and NF270 membranes. Membranes of NF70 and NF270 with surface areas of 46.2 cm2 were used throughout this research.

Four model compounds with different functional groups of benzene i.e., carboxylic and phenolic groups, were used. Phloroglucinol (1,3,5-trihroxybenzene), resorcinol (1,3-dihydroxybenzene) and 3-hydroxybenzoic acid were chosen as small THMPs, while tannic acid was chosen as the medium molar mass compound. All model compounds were purchased from Riedel-de Ha¨en (Germany).Table 2shows the properties of the model compounds andFig. 1shows the structure of the model compounds. All stock solutions were prepared using

Table 1

Characteristics of NF70 and NF270 membranes Parameter

NF70 NF270

Manufacture Dow-FilmTec Dow-FilmTec

MWCO (Da) 250 300a Materials 0.2␮m crosslinked aromatic polyamide + 0.46␮m polysulfone Semiaromatic piperzine-based aromatic polyamide layer on top of a polysulfone microporus support reinforced with a polyester non-woven backing layer CaCl2rejection (%) 85–95 40–60 MgSO4rejection (%) 98 >97 Charge (neutral pH) Negative Negative

(3)

Table 2

Physico-chemical characteristics of the model compounds used in this study Model compound

Resorcinol Phloroglucinol 3-Hydroxybenzoic acid Tannic acid

Molecular formula C6H6O2 C6H6O3 C7H6O3 C76H52O46

CAS registry number 108-46-3 108-73-6 99-06-9 NA

Molecular weight 110.11 126.11 138.12 1701.24

Dissociation constant (pKa)f pK1: 9.30; pK2: 11.06 pK1: 7.97; pK2: 9.23; pK3: 14 pK1: 3.96; pK2: 9.61 NAa

Diffusion coefficient (10−6cm2/s)b 9.934 9.631 8.853 1.968

Stokes radius (nm) 0.292 0.300 0.323 0.975

Equivalent molar radius (nm) 0.343 0.349 0.366 0.858

Log Kow 1.03c 0.55c 1.39c 7.37d

SUVA254(L/mg m) 0.47 0.67 0.73 11.8

THMFPe_{/DOC (}_␮g/mg) ₃₇₇ ₃₃₀ ₇₀ ₅₇

a_{Not available.}

b _{Calculated from the method proposed by Wilke and Chang}_[40]_. c_{Obtained from}_[42]_.

d _{Obtained from}_[43]_.

e_{Seven-day formation potential measured in this research.} f_{Dissociation constant (pKa)}_[44]_.

organic-free and de-ionized water (Milli-Q SP). The DOC (dis-solved organic matter) concentrations for the model compounds were prepared at 5.0 mg/L. The pHs of the feed solutions ranged from 3 to 10 were adjusted by 0.1 M hydrochloric acid and sodium hydroxide.

2.2. PEG and saccharide analysis

The MWCO of NF70 was estimated at 250 Da in our previous research[30]. Polyethylene glycol (PEG, Aldrich) with different molecular weights (200, 300, 400, 600 and 1500 Da) were used

(4)

to estimate the MWCO of the NF270 membrane. The DOC con-centration of each PEG solution was prepared at 20 mg/L. Feed and filtrate streams were analyzed for DOC concentrations to obtain the molecular weight distribution curves. The molecular weight of the 90% PEG retained on the membrane is taken as the MWCO of the NF270 membrane.

Three saccharides (glucose, sucrose and raffinose, Aldrich) were chosen as representative neutral solutes to estimate the pore radii of NF70 and NF270 membranes. The DOC concentration for each saccharide was prepared at 20 mg/L.

2.3. Filtration tests

A laboratory cross-flow mode filtration apparatus with a flat-sheet membrane cell was used for the filtration tests.Fig. 2is a schematic diagram of the filtration module used in this study. Fresh membranes were precompacted at the standard operating pressure of 4.80 bar (70 psi) with de-ionized water. Except for the

saccharide retention tests, all experiments of the transmembrane pressure were maintained at 4.80 bar with the temperature at 25◦C and cross-flow velocity at 0.30 m/s. For saccharide reten-tion tests, the reflecreten-tion coefficients were calculated using the retention data as a function of the water flux through the mem-brane, with operating pressures ranging from 3.4 to 11.0 bar and a temperature fixed at 25◦C. All experiments were carried out in a recycle mode with both permeate and retentate recy-cling back into the reservoir. Filtration of experimental solution containing model compounds was carried out for 24 h to avoid overestimation of rejection[31].

The retention factor, R (%) was defined as

R (%) = 1−Cp Cf × 100 (1)

where Cpis the concentration of permeate and Cfis the average

concentration of feed and concentrate.

(5)

2.4. Streaming potential measurements

The streaming potential (E) of a clean membrane was measured to determine the zeta potential (ζ) on the membrane surface, which can be used to represent the membrane surface charge. The experimental set-up was almost identical to[32]. The dimensions of the membrane samples are 76 mm× 26 mm. A 0.001 M KCl solution and a pair of platina electrodes were used to measure the streaming potential, and the pH was adjusted by adding 4 M NaOH and 4 M HCl to the solution.

The relationship between E and ζ is given by the Helmholtz–Smoluchowski equation with the Fairbrother and Mastin approach[33]:

ξ = EηκRel_PεRel,s (2)

whereP is the applied pressure, ε the dielectric permittivity, η the viscosity of the solution, Relthe electrical resistance of the

electrolyte solution, and Rel,sis the electrical resistance when the

measurement cell was filled with solution with known specific conductivityκ.

2.5. Analytical methods

THMs was analyzed after the chlorination experiment. Chlo-rination was performed with 0.05 M phosphate buffer (pH 7.0) and sodium hyperchlorite at 25◦C, with a Cl2 residual

main-tained at 3–5 mg/L after 7 days. The reaction was quenched after 168 h with sodium thiosulphate solution. DOC, pH, and THMs were measured for water samples with model compounds, fol-lowed by the QA/QC programs set forth in Standard Methods

[34]. THMs were extracted with n-pentane, and the extracts were then analyzed by GC/ECD (HP 5890-II) with a fused silica cap-illary column (Restek Mtx-5, 30 m× 0.28 mm i.d., and 1.0 ␮m film thickness). Water samples for DOC and UV (254 nm) anal-yses were filtered through a pre-washed 0.45␮m filter and then

(6)

determined by a TOC instrument (O.I. Corp. model 700) and UV spectrophotometer (Hitachi U-2000).

Fig. 3summarizes the flow chart of experimental design in this study.

3. Results and discussion

3.1. Characterization of NF membranes

The MWCO of NF70 was estimated at 250 Da in our previous research[30]. From the PEG retention curve shown inFig. 4, the molecular weight of the PEGs retained for 90% is determined as the MWCO for the NF270 membrane. The MWCO of NF 270 was estimated at 300 Da.

According to[20], the pore size of the membrane can be evaluated using the Spiegler and Kedem membrane transport model (Eqs.(3) and (4))[19]and the SHP model (Eq.(5)) with the retention rates of uncharged saccharides:

R = σ(1 − F) (1− σF) (3) with F = exp −1− σ_P Jv (4) where R is the intrinsic rejection (%),σ the reflection coefficient (%), P the solute permeability (m/s), and Jvis the permeate water

flux (L/m2h): σ = 1 − 1+16r 2 s 9r2 p 1−rs rp 2 2− 1−rs rp 2 (5) where rsis the Stokes radius of the compound (nm), and rpis

the membrane pore radius (nm).

Table 3is the estimated pore radii of NF70 and NF270 mem-branes calculated for each saccharide solute. The average pore radius of the NF70 membrane is calculated at 0.48 nm, and the average pore radius of the NF270 membrane is calculated

Fig. 4. PEGs retention curve by the NF270 membrane.

Table 3

Estimated pore radii of NF70 and NF270 membranes obtained from saccharide experiments

Compound Stokes radius (nm) Estimated pore radius (nm) NF70 NF270

Glucose 0.365 0.41 0.43

Sucrose 0.471 0.44 0.54

Raffinose 0.584 0.60 0.65

Average – 0.48 0.54

at 0.54 nm. Comparable results of estimating the average pore radii for NF70 and NF270 membranes were reported by other researchers[33–37]. These results suggest that the NF270 mem-brane possesses a larger pore size than the NF70 memmem-brane, which is consistent with its lower salt rejection shown inTable 1.

Fig. 5shows the zeta potential for the NF70 and NF270 mem-branes measured at different levels of pH value. Both memmem-branes show negative zeta potential values at neutral pH. The shape of the zeta potential curves indicates amphoteric surfaces; the higher pH values, the higher zeta potentials values. Ionizable carboxyl and amine functional groups provide the amphoteric surfaces resulting from the interfacial polymerization reaction for membrane synthesis with monomeric polyamide and poly-functional acyl halide[33]. The isoelectric point (point-of-zero charge) of each NF membrane is near pH 4, which means that the membrane surface is positively charged when solution pH less than 4 and negatively charged when solution pH higher than 4. Moreover, for solution pH higher than 4, the higher the pH value was held, the more negatively charged the membrane surface was observed. The results of streaming potential mea-surement are consistent with the findings reported by Nghiem et al.[13]and Childress and Elimelech[33].

3.2. Rejection properties of NF70 and NF270 membranes

Experiments were conducted with four model compounds (Table 2 and Fig. 1) in aqueous solutions at the same

(7)

Fig. 6. Flux and permeate quality of resorcinol (pH 7, concentration = 5.0 mg/L) as a function of time for NF70 and NF270 membranes.

transmembrane pressure of 4.80 bar and cross flow velocity (0.30 m/s). Figs. 6–9 present the normalized permeate flux (NPF) and permeate concentration of resorcinol, phlorogluci-nol, 3-hydrobenzoic acid, and tannic acid as a function of time via filtration through the NF70 and NF270 membranes. The NPF is calculated by the equation as follows:

NPF (%)=PFf

PF_i × 100 (6)

where PFfis the pure water flux after the membrane filtration

test with the target compounds, and PFi is the pure water flux

across a new membrane (m3/m2h) prior to the filtration test. The average clean water permeability of the tested membranes is 6.72 L h−1m−2bar−1for NF70 and 14.79 L h−1m−2bar−1 for NF270.

Fig. 7. Flux and permeate quality of phloroglucinol (pH 7, concentra-tion = 5.0 mg/L) as a funcconcentra-tion of time for NF70 and NF270 membranes.

Fig. 8. Flux and permeate quality of 3-hydrobenzoic acid (pH 7, concentra-tion = 5.0 mg/L) as a funcconcentra-tion of time for NF70 and NF270 membranes.

In all cases, the permeate fluxes decrease gradually in the first 10 h and then reach equilibrium in the presence of model compounds (Figs. 6–9). The permeate flux of NF70 membrane decline more rapidly than that of the NF270 membrane; the NF70 system reaches 90% of the initial capacity after 20, 6.5 and 1.8 h for phloroglucinol, 3-hydroxybenzoic acid and tannic acid, respectively. For NF270 membrane filtration tests, three small THMPs remain NPF over 90% in the filtration period, except for the tannic acid with NPF less than 90% in 5 h.

DOC in the permeate increased slightly for resorcinol and phloroglucinol as time progressed. Rejections of 79.1 and 80.6% were observed for resorcinol and phloroglucinol, respectively, at the end of the NF70 filtration experiments. Rejections of 72.5 and 74.3% were observed for resorcinol and phlorogluci-nol, respectively, at the end of the NF270 filtration experiments (Figs. 6 and 7). This rejection is likely due to the adsorption of

Fig. 9. Flux and permeate quality of tannic acid (pH 7, concentration = 5.0 mg/L) as a function of time for NF70 and NF270 membranes.

(8)

Fig. 10. Comparison of THMFP rejections for model compounds by NF70 and NF270 membranes (pH 7, concentration = 5.0 mg/L).

model compounds by hydrogen bonding between the phenolic functional groups on nonionic solutes and carbonyl functional groups on polyaimde NF membranes [13,38]. The permeate concentration presents a characteristic curve with a small break-through concentration, which means that the size exclusion becomes a dominant rejection mechanism when adsorption reaches equilibrium [13]. 3-Hydrobenzoic acid has a higher compound rejection (about 86.2% for NF70 and 80.0% for NF270 membranes) compared to phloroglucinol and resorcinol, and the DOC in permeate flux remains fairly constant during the filtration period (Fig. 8). This rejection is due to the hydrol-ysis of the carboxyl group on 3-hydrobenzoic acid (pK1= 4.06,

Table 2) at pH 7. Although the MW of 3-hydrobenzoic acid is very similar to that of resorcinol and phloroglucinol, this compound carries a negative charge at neutral pH and can be excluded by the NF70 and NF270 membranes (with negative charge at neutral pH) owing to the electrostatic repulsive inter-action with the membrane surfaces[39]. It can be concluded that the influence of charge exclusion is greater than the influence of compound adsorption on compound rejection at neutral pH.

The best quality of permeate flux was observed for tannic acid with both the NF70 and NF270 membranes (Fig. 9). The removal efficiency of tannic acid is the highest among the model compounds (96.9% for the NF70 membrane and 97.6% for the NF270 membrane) owing to its large molecular size (0.858 nm) compared to the pore radius of NF70 membrane (0.48 nm) and NF270 membrane (0.54 nm). There are many phenolic func-tional groups of tannic acid (Fig. 2), which may increase the possibility of adsorption onto polyamide membrane surfaces as mentioned above and thus resulting in more flux decline compared with the other three small THMPs (Figs. 6–9).

Fig. 10shows the THMFP rejections of model compounds by NF70 and NF270 membranes. The trend is the same as the DOC rejection, i.e., the rejections of the three small THMPs in permeate for the NF70 membrane are higher than that for the NF270 membrane, and the rejection of tannic acid is the highest for both NF70 and NF270 membranes. The high permeability of the NF270 membrane is likely due to its larger pore size. For the three small THMPs which are partially rejected by the NF270 membrane, the solute flux increases with the pure water flux and results in slightly lower rejection compared with the NF70 membrane (Figs. 6–9). However, for the highly rejected tannic acid, the high permeability in NF270 filtration comes from the pure water flux so that the rejection of tannic acid is slightly higher by the NF270 than by the NF70 membrane.

From the above experiments, it can be concluded that the NF70 and NF270 membranes are effective in removing DOC and THMFP for molecules with a wide range of molecular mass. For small compounds with high THM yield, other treatment processes may be utilized together with nanofiltraiton to enhance the removal of small compounds.

3.3. Effect of pH on compound rejections

The diffusivity coefficient and Stokes radius of each model compound inTable 2were determined by the Wilke and Chang and the Stokes–Einstein equations[40]. The Stokes radii of three small THMPs are smaller than the average pore radii of the NF70 and NF270 membranes, and the Stokes radius of tannic acid is significantly larger than the average pore size of both mem-branes. Thus, it is expected that the retention of the small THMPs would be very low and the retention would be complete for tan-nic acid if the rejection mechanism was only based on the steric hindrance. From the experimental data inFigs. 6–10, it indicates that the rejections of three small THMPs are high (72.5–86.2%), which implies that there might be another controlled mechanism, such as the charge effect, affecting the rejection ratio.

To evaluate the effect of charge exclusion, feed solutions of model compounds with different pH values were applied on both NF membranes.Fig. 11shows the relationship between the solution pH values and model compounds rejections. All data shows that rejection of all model compounds increases with an

Fig. 11. Effect of pH on rejection of model compounds by NF70 and NF270 membranes (initial concentration = 5 mg/L).

(9)

increase in pH for both NF membrane. A more profound pH effect was observed for the NF270 membrane. This pH effect is due to the polyelectrolyte character of the model compounds and the negatively charged NF membrane characteristics, which is discussed below.

For the three small THMPs, at pH values lower than the ion-ization constant of each compound, the rejection mechanism is based on the sieve effect and the adsorption of model com-pounds on NF membranes, as mentioned above. Therefore, the rejections of the three small THMPs at low pH values are sig-nificantly less than at high pH values, especially for the NF270 membrane. 3-Hydrobenzoic acid has a slightly higher compound rejection compared to phloroglucinol and resorcinol, which may be attributed to its low pK1value. The low pK1value cause the

carboxyl functional groups to start hydrolysis at pH 4.06, lead-ing to an increase in electrostatic repulsive interaction with the NF70 and the NF270 membranes[39]. However, at pH 10, two phenolic functional groups of phloroglucinol and one carboxyl plus one phenolic functional group of 3-hydrobenzoic acid are dissociated. At high pH, the charge density and molecular size of phloglucinol are similar to those of 3-hydrobenzoic acid. There-fore, the rejection efficiencies of the two compounds are similar in both NF membranes.

For tannic acid with medium MW at pH values lower than its ionization constant, the molecule is neutral and may form a compact coiled conformation because of the suppression of dis-sociation of ionogenic groups, the same as humic acid and fulvic acid studied by[41], which leads to a restriction of electrostatic intra-chain repulsion and causes a reduction in the tannic acid molecular size. Therefore, the rejection mechanism is mainly due to the “sieve effect” under acidic conditions. However, at pH values higher than the ionization constant of tannic acid, the molecule unfolds because of the ionization of ionogenic groups leading to electrostatic repulsion between macromolecule seg-ments. At the higher pH value, the tannic acid molecule is negatively charged with a larger molecular size which can be rejected more effectively by both the “sieve effect” and the “charge effect” on the NF membranes. As a result, the removal efficiency of tannic acid increases from 93.6% up to 98.5% fil-tered by the NF70 membrane and from 87.6 to 99.9% filfil-tered by the NF270 membrane.

There is a notable trend in the experimental data of NF270 membrane that the rejections of all model compounds show strong correlation with the pH values (Fig. 11), which implies that the electrostatic interaction between the NF270 membrane surface and the model compounds grows stronger as the solu-tion pH rises. The electrostatic interacsolu-tion between membrane surface and model compounds is confirmed by the results of the zeta potential measurements at different pH values (Fig. 5).

Since the membranes are more negatively charged as pH val-ues higher above the isoelecric point, the electrostatic repulsion force between the membrane surface and the model compounds increases as pH rises. Dissociation of the functional groups on the membrane and the model compounds intensifies the electro-static interaction between the NF70 and/or NF270 membranes and model compounds. As a result, the model compounds rejec-tions should increase as pH rises. The trend of the experimental

data shown inFig. 11indicates that the NF70 membrane is less pH sensitive than the NF270 membrane (Fig. 5). This implies that the solution pH would greatly affect the performance of NF270 membrane; therefore, the application of NF270 at differ-ent pH values should be evaluated and adjusted to achieve the optimum filtration performance. On the other hand, as shown inFigs. 6–9, the NF270 membrane (with average clean water permeability of 14.79 L h−1m−2bar−1) exhibits the superior permeation rate than the NF70 membrane (with average clean water permeability of 6.72 L h−1m−2bar−1). It suggests that the two-fold increase in the NF270 membrane productivity as compared with the NF70 membrane would dramatically reduce energy consumption at the same operating conditions.

4. Conclusions

The rejections of three small THMPs (resorcinol, phloroglu-cinol and 3-hydrobenzoic acid) as well as tannic acid with a wide range of molecular mass by the NF70 and NF270 membranes are effective. Filtered solutes with MWs less than the MWCO of membranes or with Stokes radii less than the membrane pore radius can be rejected satisfactorily. The compound rejection of 3-hrdrobenzoic acid is the highest due to the electrostatic repulsive interaction with the nanofiltration membranes and its ionized functional groups.

Model compounds rejection increases with pH in the solu-tion due to the dissociasolu-tion of ionizable funcsolu-tional groups on the model compounds and the electric interaction between the compounds and membranes. High removals of THMPs were observed at high pH values. The charge exclusion is found to be the prevailing mechanism of small model compounds retention by the negatively charged nanofiltration membranes. The size exclusion mechanism also is observed to be important but not sufficient for the unionized small organic molecules rejections. The NF270 membrane is more pH sensitive than the NF70 membrane. This sensitivity implies that the solution pH would greatly affect the performance of the NF270 membrane. Further research should be focused on the application of NF270 at dif-ferent pH values, and the optimum filtration performance should be determined.

In general, the NF270 membrane exhibits the superior perme-ation rate, which takes an advantage over the NF70 membrane from the aspect of energy conservation.

References

[1] C.I. Chaidou, V.I. Georgakilas, C. Stalikas, M. Saraci, E.S. Lahaniatis, Formation of chloroform by aqueous chlorination of organic compounds, Chemosphere 39 (1999) 587–594.

[2] Health Canada, Guidelines for Canadian drinking water quality: guideline technical document: trihalomethanes, in: Federal-Provincial-Territorial Committee on health and the Environment, Ottawa, Ontario, 2006. [3] US EPA, 2004 Edition of the Drinking Water Standards and health

Advi-sories, EPA 822-R-04-005, Office of Water, U.S. Environmental Protection Agency, Washington, DC, 2004.

[4] WHO, Guidelines for Drinking-Water Quality, 3rd ed., 2003.

[5] National Health and Medical Research Council, “Australian Drinking Water Guidelines”, Agricultural and Resource Management Council of Australia and New Zealand, Commonwealth of Australia, 1996.

(10)

[6] Environmental Health Directorate, Summary of Guidelines for Canadian Drinking Water Quality. Prepared by the Federal-Provincial Subcommittee on Drinking Water of the Federal-Provincial Committee of Environmental and Occupational Health, April 2003.

[7] Ministry of Health, NZ, Drinking-Water Standards for New Zealand, 2000. [8] USEPA, National primary drinking water regulations: stage 2 disinfectants and disinfection byproducts rule (stage 2 DBPR), final rule, Federal Reg. 71 (2006) 387–493.

[9] US EPA, N-nitrosodimethylamine CASRN 62-75-9, Integrated Risk Infor-mation Service (IRIS) Substance File, 1997.

[10] J.A. Nilson, F.A. Digiano, Influence of NOM composition on nanofiltration, J. Am. Water Works Assoc. 88 (1996) 53–66.

[11] D.M. Owen, G.L. Amy, Z.K. Chowdhury, R. Paode, G. McCoy, K. Vis-cosil, NOM characterization and treatability, J. Am. Water Works Assoc. 87 (1995) 46–63.

[12] T. Wintgens, M. Gallenkemper, T. Melin, Endocrine disrupter removal from wastewater using membrane bioreactor and nanofiltration technology, Desalination 146 (2002) 387–391.

[13] L.D. Nghiem, A.I. Schafer, M. Elimelech, Removal of natural hormones by nanofiltration membranes: measurement, modeling, and mechanics, Envi-ron. Sci. Technol. 38 (2004) 1888–1896.

[14] T.U. Kim, G. Amy, J.E. Drewes, Rejection of trace organic compounds by high-pressure membranes, Water Sci. Technol. 51 (2005) 335–344. [15] M.R. Wiesner, J. Hackney, S. Sethi, J.G. Jacangelo, J.M. Laine, Cost

esti-mates for membrane filtration and conventional treatment, J. Am. Water Works Assoc. 86 (1994) 33–41.

[16] J.G. Jacangelo, J. Demarco, D.M. Owen, S.J. Randtke, Selected processes for removing NOM—an overview, J. Am. Water Works Assoc. 87 (1995) 64–77.

[17] B. van der Bruggen, J. Schaep, W. Maes, D. Wilms, C. Vandecasteele, Nanofiltration as a treatment method for the removal of pesticides from ground water, Desalination 117 (1998) 139–147.

[18] Y. Kiso, Y. Nishimura, T. Kitao, K. Nishimura, Rejection of nonphenylic pesticides with nanofiltration membranes, J. Membr. Sci. 171 (2000) 229–237.

[19] K.S. Spiegler, O. Kedem, Thermodynamics of hyperfiltration (reverse osmosis): criteria for efficient membranes, Desalination 1 (1966) 311–326. [20] X.L. Wang, T. Tsuru, S. Nakao, S. Kimura, Electrolyte transport through nanofiltraiton membranes by the space-charge model and the compar-ison with Teorell–Mayer–Sievers model, J. Membr. Sci. 103 (1995) 117–133.

[21] W.R. Bowen, H. Mukhtar, Characterization and prediction of separation performance of nanofiltration membranes, J. Membr. Sci. 112 (1996) 263–274.

[22] E.E. Chang, P.C. Chaing, Y.W. Ko, W.S. Lan, Characteristics of organic pre-cursors and their relationship with disinfection by-products, Chemosphere 44 (2001) 1231–1236.

[23] P.C. Chiang, E.E. Chang, C.H. Liang, NOM characteristics and treatabili-ties of ozonation processes, Chemosphere 46 (2002) 929–936.

[24] C.W.K. Chow, R. Fabris, M. Drikas, A rapid fractionation technique to characterize natural organic matter for the optimization of water treatment processes, J. Water Supply Res. Technol.-Aqua 53 (2004) 85–92.

[25] G.L. Amy, B.C. Alleman, C.B. Cluff, Removal of dissolved organic-matter by nanofiltration, J. Environ. Eng.-ASCE 116 (1990) 200–205.

[26] D.A. Reckhow, P.C. Singer, R.L. Malcolm, Chlorination of humic materials—by-product formation and chemical interpretations, Environ. Sci. Technol. 24 (1990) 1655–1664.

[27] D.L. Norwood, J.D. Johnson, R.F. Christman, J.R. Hass, M.J. Bobenri-eth, Reactions of chlorine with selected aromatic models of aquatic humic material, Environ. Sci. Technol. 14 (1980) 187–190.

[28] G.V. Korshin, C.W. Li, M.M. Benjamin, Monitoring the properties of nat-ural organic matter through UV spectroscopy: a consistent theory, Water Res. 31 (1997) 1787–1795.

[29] IRCHS,http://www.ecn.purdue.edu/CMTI/IRCHS/.

[30] Y.L. Lin, P.C. Chiang, E.E. Chang, Reduction of disinfection by-products precursors by nanofiltration process, J. Hazard. Mater. 137 (2006) 324–331. [31] K. Kimura, S. Toshima, G. Amy, Y. Watanabe, Rejection of neutral endocrine disrupting compounds (EDCs) and pharmaceutical active com-pounds (PhACs) by RO membranes, J. Membr. Sci. 245 (2004) 71–78. [32] J.M.M. Peeters, M.H.V. Mulder, H. Strathmann, Streaming potential

mea-surements as a characterization method for nanofiltration membranes, Colloids Surf. A: Physicochem. Eng. Aspects 150 (1999) 247–259. [33] A.E. Childress, M. Elimelech, Effect of solution chemistry on the

sur-face charge of polymeric reverse osmosis and nanofiltration membranes, J. Membr. Sci. 119 (1996) 253–268.

[34] APHA Standard Methods for the Examination of Water and Wastewater, 20th ed., American Public Health Association, Washington, DC, 1998. [35] B. van der Bruggen, J. Schaep, D. Wilms, C. Vandecasteele, Influence of

molecular size, polarity and charge on the retention of organic molecules by nanofiltration, J. Membr. Sci. 156 (1999) 29–41.

[36] B. van der Bruggen, B. Daems, D. Wilms, C. Vandecasteele, Mechanisms of retention and flux decline for the nanofiltration of dye baths from the textile industry, Sep. Purif. Technol. 22–23 (2001) 519–528.

[37] N. Hilal, H. Al-Zoubi, N.A. Darwish, A.W. Mohammad, Characterization of nanofiltraiton membranes using aromatic force microscopy, Desalination 177 (2005) 187–199.

[38] M.E. Williams, J.A. Hestekin, C.N. Smothers, D. Bhattacharyya, Separa-tion of organic pollutants by reverse osmosis and nanofiltraSepara-tion membranes: mathematical models and experimental verification, Ind. Eng. Chem. Res. 38 (1999) 3683–3695.

[39] Y.Z. Xu, R.E. Lebrun, Investigation of the solute separation by charged nanofiltration membrane: effect of pH, ionic strength and solute type, J. Membr. Sci. 158 (1999) 93–104.

[40] C.J. Geankoplis, Transport Processes and Unit Operations, 3rd ed., Prentice-Hall, Englewood Cliff, NJ, 1993.

[41] M. Kabsch-Korbutowicz, K. Majewska-Nowak, T. Winnicki, Analysis of membrane fouling in the treatment of water solutions containing humic acids and mineral salts, Desalination 126 (1999) 179–185.

[42] Syracuse Research Corporation Interactive PhysProp Database Demo, 2004.

[43] Virtual Computational Chemistry Laboratory, ALOGPS 2.1 Program, 2006.

[44] Serjeant, Dempsey, Ionisation Constants of Organic Acids in Aqueous Solution, IUPAC Chemical Data Series, No. 23, Pergamon Press, 1979.